A lesson from Foodpro 2026: one wrong word on a label — "contains" where it should read "may be present" — can trigger a recall that costs millions. We're exploring an AI-assisted label review tool for the Australian food supply chain, and right now we're just trying to understand the problem properly.
Under the FSANZ Food Standards Code, allergens actually present in a product must carry a mandatory "Contains" statement, while "may be present" is a voluntary precautionary warning. Confuse the two — or get the weight, nutrition panel or a claim wrong — and the fix isn't an edit. It's a recall. And a recall reaches everyone in the chain: the manufacturer, the importer, the private-label brand, the distributor.
"Contains nuts, milk" versus "may contain nuts, milk" are not interchangeable. Under-declaring an allergen — or using a precautionary warning where a mandatory one is required — is one of the most common recall triggers in Australia.
Wrong net weight, an out-of-spec nutrition panel, or a claim that isn't permitted. Small errors that fail at inspection, get shipments rejected, or force a relabel across an entire production run.
Once it's on shelf, the options are all costly: pull the product, repackage it, or — when repackaging isn't worth it — destroy it. For a full run, the loss can reach into the millions.
Illustrative only — a concept preview, not a working product. It shows the kind of check we're exploring: label artwork read against the FSANZ Code, with issues flagged before print sign-off.
You produce and print. You've caught label errors before a run — or worse, after. We want to understand where they slip through, how sign-off works today, and what a wrong label actually costs you.
You put your name on product made elsewhere and carry the compliance risk. We want to understand how you check artwork against the FSANZ Code — and where that process is slow or uncertain.
You're the last line before a label reaches shelf. We want to understand what you check, how long it takes, and what you wish you could catch automatically.
We're not selling anything. We're trying to understand whether label errors hurt often enough, and cost enough, to be worth solving properly. If you work in food manufacturing, importing, private label or QA — we'd genuinely like your perspective.
This is a concept in discovery phase — not a finished product, and not compliance advice. Nothing shared in conversations will be published or attributed without your permission.